Assistants, knowledge tools and domain platforms
Enterprise AI tends to arrive in layers. A general assistant gives every employee drafting, summarising and analysis. Knowledge tools answer questions from internal documents and should surface only what each user may already open. Automation and agents act inside business systems. Domain platforms serve one discipline in depth, such as software development, customer service or content. The layers have different owners and different tests: the assistant usually belongs to IT and is judged on adoption, while a content platform belongs to marketing and is judged on what reaches customers. Marketing fits into this stack as one domain among several, so it should build on the enterprise's AI policies rather than start from a blank page.
Why a general assistant is not a content studio
Assistants are built around a conversation with one person. Brand rules live in whatever instructions that person typed, so ten users hold ten slightly different versions of the brand. Images come back redrawn rather than reproduced, with no way to lock a real product. There is no project where a designer, a copywriter and a market reviewer work on the same set, no sign-off bound to a version, and no export in each channel's formats. Outputs sit in a chat history instead of the DAM, without a record of the model and inputs behind them. For drafts, research and internal copy, an assistant is fine. For published brand assets at volume, it is the wrong tool.
One enterprise baseline, plus what content adds
Many enterprises set their AI baseline when the general assistant went through review: single sign-on over SAML or OIDC, SCIM provisioning, data residency, zero retention, no training on inputs, audit logs and control over which models are used. A content platform should meet that baseline in full, then cover what the assistant review never asked: whether the image, video and voice models are commercially safe, how consent works for a likeness or a cloned voice, whether outputs carry machine-readable marking, and which model made each published asset. On that baseline, Synthetic White offers SSO through Google and Microsoft, SCIM, EU or US data residency, zero-retention agreements with every model provider and confidential work pinned to self-hosted models.
EU AI Act duties that land on the deployer
A company using a content platform is normally a deployer under the EU AI Act, and its own Article 50 duties have applied since 2 August 2026. It must disclose deep fakes, which the Commission reads to include people or scenes that could plausibly exist, not only real ones. It must also disclose AI-generated text published to inform the public on matters of public interest, unless the text has had human review or editorial control and a person or organisation holds editorial responsibility. Machine-readable marking is the providers' job, with generative systems already on the market given until 2 December 2026. Article 4, as amended in 2026, requires providers and deployers to take measures to support the AI literacy of staff and others using AI systems on their behalf, without mandating a set level.
Updated 25 September 2026